Hydrological model parameter optimization method and system based on genetic algorithm

Abstract

The present application relates to the technical field of genetic algorithm, specifically to a hydrological model parameter optimization method and system based on genetic algorithm, in the present application, by means of low contribution vector marking, comparison of exploration group candidate error and main group quantile error, pairing of immigration items and emigration items, parameter combinations in the low error contribution area for a long time and close to the elite vector area are replaced, beneficial effects such as genetic main group shrinkage relief, chromosome difference reservation and cross-region candidate parameters entering the main group are obtained, by means of stagnation generation accumulation, parameter discrete value comparison, feasible proportion calculation and simulated annealing algorithm screening of deteriorated candidate vectors, candidate vectors with slightly higher error but larger parameter distance are reserved to enter the subsequent sequence, beneficial effects such as enhanced ability to jump out of local extreme value, reduced probability of premature convergence and maintained search activity in the later stage are obtained, and the hydrological model optimization parameter group is not dominated by a single error index, beneficial effects such as runoff process matching, water balance, flood peak response and change of the water recession section are obtained.

Description

Technical Field

[0001] This invention relates to the field of genetic algorithm technology, and in particular to a method and system for optimizing hydrological model parameters based on genetic algorithms. Background Technology

[0002] The field of genetic algorithm technology aims to utilize population evolution search mechanisms to determine, within a limited parameter search space, the parameter combinations that enable the evaluation error between the hydrological model calculation results and the target reference results to meet the set optimization conditions, under the conditions of multi-parameter coupling in hydrological models and nonlinear changes in the objective function.

[0003] The purpose of the genetic algorithm-based hydrological model parameter optimization method is to address the problems of coupling relationships among multiple parameters to be calibrated in hydrological models, large parameter search space, and nonlinear characteristics of the objective function. By performing selection, crossover, mutation, and iterative updates on candidate parameter combinations through a genetic algorithm, the method determines the parameter combinations within a limited parameter range that satisfy the set evaluation conditions between the hydrological model calculation results and the target reference results. This reduces the probability of getting trapped in local extrema during the local search and improves the efficiency of hydrological model parameter calibration and the consistency between the model calculation results and the target reference results.

[0004] Existing technologies typically employ a single candidate population for selection, crossover, mutation, and iterative updates. Under the same selection pressure, candidate parameter combinations continuously concentrate towards the current low-error region. When runoff parameters, infiltration parameters, and groundwater runoff parameters are coupled, the early dominance of a certain parameter locus will cause the associated parameter locus to shrink synchronously, leading to a decrease in chromosome differences and a gradual narrowing of the search range with increasing iterations. If the flood peak deviation decreases rapidly while the water volume deviation remains high, simply sorting by total error will retain the parameter combination with the best flood peak fit. Evapotranspiration and interflow parameters may remain in localized areas for extended periods, making it difficult for the error in the receding water stage to decrease synchronously. Using the same perturbation scale for fixed crossover and mutation probabilities in both the early and later iterations can result in a large number of out-of-bounds vectors in the early stage and a lack of cross-regional candidate combinations in the later stage. For candidate vectors with slightly higher errors but larger parameter distances, existing processes often directly eliminate them based on fitness sorting, preventing potentially low-error regions in the parameter space from entering subsequent iterations. This makes the hydrological model calibration results susceptible to local extrema, single-term error weighting, and premature convergence. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by proposing a method and system for optimizing hydrological model parameters based on genetic algorithms.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for optimizing hydrological model parameters based on genetic algorithms, comprising the following steps: S1: Based on runoff generation parameters, evapotranspiration parameters, infiltration parameters, interflow parameters, groundwater runoff parameters, confluence parameters, upper limit of parameters, and lower limit of parameters, bind chromosome position numbers, compare boundaries, remove out-of-bounds vectors, divide the genetic main group and the exploration group, and obtain the hydrological parameter boundary group table; S2: Based on the hydrological parameter boundary group table, a genetic algorithm is used to sort the candidate vector errors, select the window winning vector, exchange parent sites, limit the perturbation amplitude of offspring, mark low contribution vectors, and pair immigration and emigration items to obtain a two-group conditional migration table. S3: Based on the bigroup conditional migration table, replace the migration items, retain the elite vectors, accumulate the stagnation algebras, compare the discrete values ​​of the parameters, calculate the feasible proportion, use the simulated annealing algorithm to screen the deterioration candidate vectors, and obtain the search state operator table. S4: Based on the search state operator table, check the boundaries and parameter constraints of the temporary retained items, remove the abnormal vectors, calculate the runoff process error, water volume deviation, flood peak deviation, and drainage section deviation, and obtain the true error candidate sequence; S5: Based on the real error candidate sequence, call the search state operator table, compare the upper limit algebra, the unimproved algebra, and the target error threshold, select and map six types of hydrological parameters according to the error order, and obtain the hydrological model optimization parameter set.

[0007] As a further embodiment of the present invention, the hydrological parameter boundary group table includes parameter name, chromosome position number, upper limit value of parameter, lower limit value of parameter, and group number; the bi-group conditional migration table includes candidate vector number, error sequence, low contribution vector number, inflow item number, and outflow item number; the search state operator table includes elite vector number, stagnation algebra, parameter discrete value, feasible proportion value, and deterioration candidate vector number; the true error candidate sequence includes candidate vector number, runoff process error value, water volume deviation value, flood peak deviation value, and recession section deviation value; and the hydrological model optimization parameter set includes runoff generation parameter value, evapotranspiration parameter value, infiltration parameter value, interflow parameter value, groundwater runoff parameter value, and confluence parameter value.

[0008] As a further aspect of the present invention, the specific steps for obtaining the hydrological parameter boundary group table are as follows: Based on runoff generation parameters, evapotranspiration parameters, infiltration parameters, interflow parameters, groundwater runoff parameters, confluence parameters, upper limit of parameters, and lower limit of parameters, the six types of parameters are arranged in order, the parameter values ​​are compared with the boundary values, out-of-bounds markers are written, and a boundary verification vector set is generated. Based on the boundary verification vector set, out-of-bounds markers are filtered, reserved vector chromosome positions are bound, group numbers are assigned, and genetic master group numbers and exploration group numbers are written to obtain the hydrological parameter boundary group table.

[0009] As a further aspect of the present invention, the specific steps for obtaining the bigroup conditional migration table are as follows: Based on the hydrological parameter boundary group table, a genetic algorithm is used to read the candidate vector number of the genetic main group, the six types of parameter numbers, and the error value. The candidate vectors are arranged in ascending order of error, the same window comparison interval is defined, and the vector with lower error is selected window by window and the parent source number is recorded to generate the window winning vector set. Based on the windowed winning vector set, the parent source number and six types of parameter numbers are read, the same-position parameter segments are swapped, the perturbation amplitude of the offspring is limited, the upper and lower limits of parameters are compared item by item, the offspring vectors that exceed the limits are eliminated and the qualified vectors of the boundary are retained to generate the offspring candidate vector set. Based on the candidate vector set of offspring, read the candidate vector error order, the number of consecutive low positions, and the elite vector distance value, mark the low contribution vectors, compare the candidate error of the exploration group with the quantile error of the main group, select the immigration item number and the immigration item number, and establish a bi-group conditional migration table.

[0010] As a further aspect of the present invention, the genetic algorithm is employed to read the candidate vector number of the genetic main population, the six types of parameter positions, the error value, the upper limit of the parameter, and the lower limit of the parameter. The candidate vectors of the genetic main population are arranged in ascending order of error. A comparison interval is defined for each of the two to five candidate vectors. The runoff process error, water volume deviation, flood peak deviation, and recession section deviation are compared window by window. The candidate vector with the lowest error value is selected as the parent vector. The parent source number, parent group number, and parent chromosome position number are recorded. Parent isotopic parameter segments are exchanged according to the six types of parameter positions. The perturbation amplitude of the offspring is limited by the reciprocal of the parameter dimension. The offspring parameter values, upper limit of the parameter, and lower limit of the parameter are compared item by item. Off-limit offspring vectors are deleted, and qualified offspring vectors are retained. The parent source number, offspring vector number, error sequence, and boundary verification mark are summarized to generate a set of offspring candidate vectors.

[0011] As a further aspect of the present invention, the specific steps for obtaining the search state operator table are as follows: Based on the bi-group conditional migration table, the immigration item number, immigration item number, main group position number, and elite vector number are read. The immigration item is written into the immigration item position, the elite vector number and the corresponding six types of parameter values ​​are retained, the order of the genetic main group candidate vectors is rearranged, and the main group update vector set is generated. Based on the main group update vector set, read the continuous generation target error, the distribution values ​​of six types of parameters, the number of feasible vectors, the cumulative number of stagnant generations, calculate the parameter discrete values, calculate the feasible ratio values, write the search status markers according to the threshold interval, and generate a search status determination table. Based on the search state determination table, the simulated annealing algorithm is used to read the deterioration candidate vector number, parent error value, candidate error value, temperature value, and random number. The error difference is compared with the retention threshold, unqualified deterioration vectors are screened out, and the retention vector numbers are registered to obtain the search state operator table.

[0012] As a further aspect of the present invention, the simulated annealing algorithm is used to read the deterioration candidate vector number, parent error value, candidate error value, temperature value, random number, current generation number, and temperature segment number. The error difference is calculated by subtracting the parent error value from the candidate error value. When the candidate error value is less than the parent error value, the retained vector number is registered. When the candidate error value is greater than the parent error value, the exponential ratio between the error difference and the temperature value is calculated. The exponential ratio is used as the retention threshold. The random number is compared with the retention threshold. When the random number is less than the retention threshold, the retained vector number is registered. When the random number is greater than the retention threshold, the unqualified deteriorated vector is marked. The unqualified deteriorated vectors are filtered out. The retained vector number, temperature segment number, error difference, and retention threshold are summarized to obtain the search state operator table.

[0013] As a further aspect of the present invention, the specific steps for generating the true error candidate sequence are as follows: Based on the search state operator table, read the temporary retained item number, check the parameter boundary and the constraints between parameters, mark the out-of-bounds vector and the constraint conflict vector, and generate a temporary vector verification table. Based on the temporary vector verification table, abnormal vectors are removed, and runoff process error, water volume deviation, flood peak deviation, and receding section deviation are calculated. Candidate positions are arranged to obtain the true error candidate sequence.

[0014] As a further aspect of the present invention, the specific steps for obtaining the hydrological model optimization parameter set are as follows: Based on the real error candidate sequence, the search state operator table is called, the upper limit algebra, the unimproved algebra, and the target error threshold are compared, a termination flag is written, and a termination judgment sequence table is generated. Based on the termination determination sequence list, the error sequence of runoff process, the sequence of water volume deviation, the sequence of flood peak deviation, and the sequence of water receding section deviation are read, and six types of parameter values ​​are mapped to obtain the hydrological model optimization parameter set.

[0015] A system for optimizing hydrological model parameters based on a genetic algorithm, wherein the system is used to execute the above-mentioned method for optimizing hydrological model parameters based on a genetic algorithm, the system comprising: Encoding and grouping module: Based on runoff generation parameters, evapotranspiration parameters, infiltration parameters, interflow parameters, groundwater runoff parameters, confluence parameters, upper limit of parameters, and lower limit of parameters, the module arranges the parameter positions, binds the chromosome positions, compares the boundaries, deletes out-of-bounds vectors, and obtains the hydrological parameter boundary group table; Interactive scheduling module: Based on the hydrological parameter boundary group table, a genetic algorithm is used to sort candidate errors, select window winning vectors, exchange parent sites, limit the perturbation amplitude of offspring, mark low contribution vectors, and pair immigration and emigration items to obtain a dual-group conditional migration table. Annealing screening module: Based on the bigroup conditional migration table, replace the migration items, retain the elite vectors, accumulate the stagnation algebras, compare the discrete values ​​of parameters, calculate the feasible proportion, and use the simulated annealing algorithm to screen the deteriorated candidate vectors to obtain the search state operator table; Runoff accounting module: Based on the search state operator table, it checks the boundaries and parameter constraints of temporary retained items, deletes abnormal vectors, calculates runoff process error, water volume deviation, flood peak deviation, and drainage section deviation, and obtains the true error candidate sequence; Termination output module: Based on the real error candidate sequence, call the search state operator table, compare the upper limit algebra, the unimproved algebra, and the target error threshold, map six types of parameters according to the error order, and obtain the hydrological model optimization parameter set.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, by sorting candidate vector errors, selecting window winning vectors, exchanging parental loci, and limiting the perturbation amplitude of offspring, the dominant parameter fragments within the genetic main population continuously participate in the iteration, thereby achieving the beneficial effect of stable updating of coupled parameter combinations such as runoff generation parameters, infiltration parameters, and groundwater runoff parameters. In this invention, by marking low contribution vectors, comparing the candidate error of the exploratory group with the quantile error of the main group, and pairing immigration terms with immigration terms, the parameter combinations that have been in the low error contribution region and are close to the elite vector region for a long time are replaced, thereby obtaining beneficial effects such as alleviating the contraction of the genetic main group, preserving chromosome differences, and allowing cross-regional candidate parameters to enter the main group. In this invention, by stagnant algebraic accumulation, parameter discrete value comparison, feasible proportional calculation, and simulated annealing algorithm to screen deterioration candidate vectors, candidate vectors with slightly higher errors but larger parameter distances are retained in the subsequent sequence. This achieves beneficial effects such as enhanced local extremum escape capability, reduced premature convergence probability, and maintained search activity in the later stage. It also makes the hydrological model's parameter set optimization not dominated by a single error index, and achieves beneficial effects that take into account runoff process matching, water balance, flood peak response, and changes in the receding section. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the workflow of the present invention; Figure 2 This is a system flowchart of the present invention. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0019] Example 1 Please see Figure 1This invention provides a technical solution: a method for optimizing hydrological model parameters based on genetic algorithms, comprising the following steps: S1: Based on runoff generation parameters, evapotranspiration parameters, infiltration parameters, interflow parameters, groundwater runoff parameters, confluence parameters, upper limit of parameters, and lower limit of parameters, bind chromosome position numbers, compare boundaries, remove out-of-bounds vectors, divide the genetic main group and the exploration group, and obtain the hydrological parameter boundary group table; S2: Based on the boundary group table of hydrological parameters, a genetic algorithm is used to sort the candidate vector errors, select the winning vector of the window, exchange parent sites, limit the perturbation amplitude of offspring, mark low contribution vectors, and pair immigration and emigration items to obtain a two-group conditional migration table. S3: Based on the bigroup conditional migration table, replace the migration terms, retain the elite vectors, accumulate the stagnation algebras, compare the discrete values ​​of the parameters, calculate the feasible proportion, use the simulated annealing algorithm to screen the deterioration candidate vectors, and obtain the search state operator table. S4: Based on the search state operator table, check the boundaries and parameter constraints of the temporary retained items, remove the outlier vectors, calculate the runoff process error, water volume deviation, flood peak deviation, and drainage section deviation, and obtain the true error candidate sequence; S5: Based on the real error candidate sequence, call the search state operator table, compare the upper limit algebra, the unimproved algebra, and the target error threshold, select and map six types of hydrological parameters according to the error order, and obtain the hydrological model optimization parameter set.

[0020] The hydrological parameter boundary group table includes parameter name, chromosome number, upper limit value, lower limit value, and group number. The bi-group conditional migration table includes candidate vector number, error order, low contribution vector number, inflow item number, and outflow item number. The search state operator table includes elite vector number, stagnation algebra, parameter discrete value, feasible proportion value, and deterioration candidate vector number. The true error candidate sequence includes candidate vector number, runoff process error value, water volume deviation value, flood peak deviation value, and recession section deviation value. The hydrological model optimization parameter group includes runoff generation parameter value, evapotranspiration parameter value, infiltration parameter value, interflow parameter value, groundwater runoff parameter value, and confluence parameter value.

[0021] The specific steps for generating the boundary group table of hydrological parameters are as follows: Based on parameters such as runoff generation, evapotranspiration, infiltration, interflow, groundwater runoff, and confluence, and considering upper and lower limits of these parameters, Latin hypercube sampling and real chromosome coding are employed. The parameter dimension is set to 6, and the sample size is set to 10 times the number of parameters across the 6 categories. Parameter intervals are divided according to dimension numbers 1 to 6, with each interval further divided into sample size segments. One real value is extracted from each segment. The 6 real values ​​are then written into chromosome positions in the order of runoff generation, evapotranspiration, infiltration, interflow, groundwater runoff, and confluence. Parameter values, upper and lower limits are compared item by item. Parameter values ​​greater than the upper limit are marked as exceeding the upper limit, values ​​less than the lower limit are marked as exceeding the lower limit, and values ​​between the upper and lower limits are marked as passing. This generates a boundary verification vector set. Based on the boundary verification vector set, the upper boundary cross-boundary marker, lower boundary cross-boundary marker, qualified marker, sample number, and chromosome position are read through the boundary marker screening rule and the proportional grouping rule. Vectors with upper and lower boundary cross-boundary markers are deleted, and vectors with qualified markers are retained. 70% of the retained vectors are assigned to the genetic main group, and the remaining retained vectors are assigned to the exploration group. The genetic main group number is written as 1, and the exploration group number is written as 2. The group number, sample number, chromosome position number, six types of parameter values, parameter upper limit, and parameter lower limit are written into the same row of records to obtain the hydrological parameter boundary group table.

[0022] The specific steps for generating a bigroup conditional transition table are as follows: Based on the boundary group table of hydrological parameters, a genetic algorithm and tournament selection are used to read the candidate vector number of the main genetic group, the position numbers of 6 types of parameters, the runoff process error, the water volume deviation, the flood peak deviation, and the recession section deviation. The tournament size is set to 2 to 5 candidate vectors. The candidate vectors are arranged in ascending order of error. The comparison interval is defined by the window number. The four error values ​​are compared window by window. The candidate vector with the lowest error value is selected. The parent source number, parent group number, and parent chromosome position number are recorded to generate the window winning vector set. Based on the windowed winning vector set, by simulating binary crossover and polynomial mutation, the parent source number, six types of parameter bits, parameter upper limit, and parameter lower limit are read. The crossover probability is set to 0.7 to 0.95, and the mutation probability is set to the reciprocal of the parameter dimension. Parent parameter fragments are swapped according to the same parameter bits, and the child parameter increment is written according to the perturbation exponent. The child parameter value is calculated, and the child parameter value, parameter upper limit, and parameter lower limit are compared item by item. Off-boundary child vectors are deleted, and qualified boundary vectors are retained to generate a child candidate vector set. Based on the candidate vector set of offspring, the error order, number of consecutive low positions, elite vector distance value, exploratory group candidate error, and main group quantile error are read. The difference weight is set to 0.4 to 0.9, and the binomial crossover probability is set to 0.5 to 0.95. The difference candidate vector is calculated by randomly selecting 3 sets of parameter vectors, and the difference candidate vector is calculated by selecting excellent parameter vectors and randomly selecting 2 sets of parameter vectors. Vectors with error order in the bottom 30% for 3 to 8 consecutive generations and elite vector distance value less than the threshold vector are marked. The exploratory group candidate error is compared with the main group 25% to 40% quantile error. The immigration item number and immigration item number are selected, and a bi-group conditional migration table is established.

[0023] The genetic algorithm reads the candidate vector number of the main genetic group, the position numbers of six types of parameters, the error value, the upper limit of the parameter, and the lower limit of the parameter. It sorts the candidate vectors of the main genetic group in ascending order of error, defines a comparison interval for two to five candidate vectors, and compares the runoff process error, water volume deviation, flood peak deviation, and recession section deviation window by window. It selects the candidate vector with the lowest error value as the parent vector and records the parent source number, parent group number, and parent chromosome position number. It exchanges parent isotopic parameter segments according to the position numbers of the six types of parameters, limits the perturbation amplitude of the offspring by the reciprocal of the parameter dimension, compares the offspring parameter values, upper limit of the parameter, and lower limit of the parameter item by item, deletes offspring vectors that exceed the limit, retains offspring vectors that meet the boundary, and summarizes the parent source number, offspring vector number, error sequence, and boundary verification mark to generate the offspring candidate vector set.

[0024] The specific steps for generating the search state operator table are as follows: Based on the dual-group conditional migration table, the following parameters are read: inbound item number, outbound item number, main group location number, elite vector number, main group quantity, and 6 types of parameter values. The elite retention quantity is set to 5% to 15% of the main group quantity. The runoff generation parameter value, evapotranspiration parameter value, infiltration parameter value, interflow parameter value, groundwater runoff parameter value, and confluence parameter value are locked according to the elite vector number. The main group row number is located according to the outbound item location number. The parameter vector associated with the inbound item number is written into the main group row number. The outbound item number is written into the exploratory group candidate pool. The genetic main group candidate vectors are rearranged according to the error order to generate the main group update vector set. Based on the main group update vector set, read the target error of consecutive generations, the distribution values ​​of 6 types of parameters, the number of feasible vectors, the number of main group vectors, and the iteration generation number. Set the stagnation window to 5 to 15 generations and the feasible proportion threshold to 0.6 to 0.95. Compare the target error difference with the improvement magnitude threshold generation by generation, accumulate the number of generations that have not reached the improvement magnitude, calculate the mean variance of the 6 types of parameters as the parameter discrete value, calculate the number of feasible vectors divided by the number of main group vectors as the feasible proportion value, and write the results into the vector row according to the 4 types of labels: extended search, stable improvement, local stagnation, and convergence confirmation, and generate a search status judgment table. Based on the search state determination table, the simulated annealing algorithm is used to read the deterioration candidate vector number, parent error value, candidate error value, temperature value, random number, algebra number, and temperature segment number. The temperature value of the first 30% of algebras is set to 1.0, the temperature value of the middle 40% of algebras is set to decrease by a factor of 0.85, and the lower limit of the temperature of the last 30% of algebras is set to 0.01. The error difference is calculated by subtracting the parent error value from the candidate error value. When the error difference is less than 0, the vector number to be retained is recorded. When the error difference is greater than 0, the negative error difference is calculated and divided by the temperature value, and the exponent value is taken as the retention threshold. The random number between 0 and 1 is compared with the retention threshold to filter out unqualified deterioration vectors and record the retention vector numbers, thus obtaining the search state operator table.

[0025] The simulated annealing algorithm reads the candidate deterioration vector number, parent error value, candidate error value, temperature value, random number, current generation number, and temperature segment number. It calculates the error difference by subtracting the parent error value from the candidate error value. When the candidate error value is less than the parent error value, it registers the retained vector number. When the candidate error value is greater than the parent error value, it calculates the exponential ratio between the error difference and the temperature value and uses the exponential ratio as the retention threshold. It compares the random number with the retention threshold. When the random number is less than the retention threshold, it registers the retained vector number. When the random number is greater than the retention threshold, it marks the unqualified deteriorated vector and filters out the unqualified deteriorated vectors. It summarizes the retained vector number, temperature segment number, error difference, and retention threshold to obtain the search state operator table.

[0026] The specific steps for generating true error candidate sequences are as follows: Based on the search state operator table, and through boundary projection judgment rules and linear constraint matrix judgment rules, the following parameters are read: temporary retention item number, runoff generation parameter value, evapotranspiration parameter value, infiltration parameter value, interflow parameter value, groundwater runoff parameter value, confluence parameter value, upper limit of parameters, lower limit of parameters, constraint matrix between parameters, and constraint threshold. The upper limit difference is calculated by subtracting the upper limit of parameters from the parameter value for each parameter, and the lower limit difference is calculated by subtracting the parameter value from the lower limit of parameters for each parameter. If the upper limit difference is greater than 0, an upper limit over-boundary mark is written; if the lower limit difference is greater than 0, a lower limit over-boundary mark is written. The six types of parameter values ​​are written into the constraint matrix product as column vectors. If the constraint matrix product is greater than the constraint threshold, a constraint conflict mark is written. The temporary retention item number, over-boundary mark, and constraint conflict mark are summarized to generate a temporary vector verification table. Based on the temporary vector verification table, and through the water balance error calculation rules, flood peak relative deviation calculation rules, and receding slope difference calculation rules, candidate vector numbers, boundary crossing markers, constraint conflict markers, simulated runoff sequences, measured runoff sequences, simulated flood peak values, measured flood peak values, simulated receding slopes, and measured receding slopes are read. Vectors with upper boundary crossing markers, lower boundary crossing markers, and constraint conflict markers are deleted. The sum of squares of the time-by-time differences between the simulated runoff sequence and the measured runoff sequence is calculated as the runoff process error. The simulated total runoff minus the measured total runoff and divided by the measured total runoff is calculated as the water volume deviation. The simulated flood peak value minus the measured flood peak value and divided by the measured flood peak value is calculated as the flood peak deviation. The simulated receding slope minus the measured receding slope is calculated as the receding slope deviation. Candidate vectors are arranged according to the order of the four error items to obtain the true error candidate sequence.

[0027] The specific steps for generating the parameter set for hydrological model optimization are as follows: Based on the true error candidate sequence, the search state operator table is invoked. Using a three-condition termination decision rule and a Boolean flag writing rule, the candidate vector number, current algebra number, upper limit algebra, number of consecutive unimproved algebras, unimproved algebra threshold, target error value, target error threshold, search state flag, and retained vector number are read. The current algebra number is compared with the upper limit algebra, the number of consecutive unimproved algebras is compared with the unimproved algebra threshold, and the target error value is compared with the target error threshold. When the current algebra number is greater than or equal to the upper limit algebra, an algebra termination flag is written. When the number of consecutive unimproved algebras is greater than or equal to the unimproved algebra threshold, a stagnation termination flag is written. When the target error value is less than or equal to the target error threshold, an error termination flag is written. The candidate vector number, search state flag, retained vector number, algebra termination flag, stagnation termination flag, and error termination flag are written into the same row of records to generate a termination decision sequence table. Based on the termination decision sequence list, and through the four-level error dictionary order sorting rules and parameter tag number mapping rules, the following parameters are read: runoff process error sequence, water volume deviation sequence, flood peak deviation sequence, receding section deviation sequence, candidate vector number, runoff generation parameter tag number, evapotranspiration parameter tag number, infiltration parameter tag number, interflow parameter tag number, groundwater runoff parameter tag number, and confluence parameter tag number. First, the candidate vectors are arranged according to the runoff process error sequence, and then the candidate vectors are arranged in sequence according to the water volume deviation sequence, flood peak deviation sequence, and receding section deviation sequence. The first candidate vector is read, and the runoff generation parameter value, evapotranspiration parameter value, infiltration parameter value, interflow parameter value, groundwater runoff parameter value, and confluence parameter value are extracted according to the six types of parameter tag numbers to obtain the hydrological model optimization parameter set.

[0028] Please see Figure 2 A system for optimizing hydrological model parameters based on genetic algorithm, used to execute the above-mentioned method for optimizing hydrological model parameters based on genetic algorithm, includes: Encoding and grouping module: Based on runoff generation parameters, evapotranspiration parameters, infiltration parameters, interflow parameters, groundwater runoff parameters, confluence parameters, upper limit of parameters, and lower limit of parameters, the module arranges the parameter positions, binds the chromosome positions, compares the boundaries, deletes out-of-bounds vectors, and obtains the hydrological parameter boundary group table; Interactive scheduling module: Based on the boundary group table of hydrological parameters, a genetic algorithm is used to sort candidate errors, select the winning vector of the window, exchange parent sites, limit the perturbation amplitude of offspring, mark low contribution vectors, and pair immigration and emigration items to obtain a two-group conditional migration table. Annealing screening module: Based on the bigroup conditional migration table, it replaces the migration out items, retains the elite vectors, accumulates the stagnation algebras, compares the discrete values ​​of parameters, calculates the feasible proportion, and uses the simulated annealing algorithm to screen the deteriorated candidate vectors to obtain the search state operator table; Runoff accounting module: Based on the search state operator table, it checks the boundaries and parameter constraints of temporary retained items, deletes abnormal vectors, calculates runoff process error, water volume deviation, flood peak deviation, and drainage section deviation, and obtains the true error candidate sequence; Termination output module: Based on the real error candidate sequence, it calls the search state operator table, compares the upper limit algebra, the unimproved algebra, and the target error threshold, maps six types of parameters according to the error order, and obtains the hydrological model optimization parameter set.

[0029] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for optimizing hydrological model parameters based on genetic algorithm, characterized in that, Includes the following steps: S1: Based on runoff generation parameters, evapotranspiration parameters, infiltration parameters, interflow parameters, groundwater runoff parameters, confluence parameters, upper limit of parameters, and lower limit of parameters, bind chromosome position numbers, compare boundaries, remove out-of-bounds vectors, divide the genetic main group and the exploration group, and obtain the hydrological parameter boundary group table; S2: Based on the hydrological parameter boundary group table, a genetic algorithm is used to sort the candidate vector errors, select the window winning vector, exchange parent sites, limit the perturbation amplitude of offspring, mark low contribution vectors, and pair immigration and emigration items to obtain a two-group conditional migration table. S3: Based on the bigroup conditional migration table, replace the migration items, retain the elite vectors, accumulate the stagnation algebras, compare the discrete values ​​of the parameters, calculate the feasible proportion, use the simulated annealing algorithm to screen the deterioration candidate vectors, and obtain the search state operator table. S4: Based on the search state operator table, check the boundaries and parameter constraints of the temporary retained items, remove the abnormal vectors, calculate the runoff process error, water volume deviation, flood peak deviation, and drainage section deviation, and obtain the true error candidate sequence; S5: Based on the real error candidate sequence, call the search state operator table, compare the upper limit algebra, the unimproved algebra, and the target error threshold, select and map six types of hydrological parameters according to the error order, and obtain the hydrological model optimization parameter set.

2. The method for optimizing hydrological model parameters based on genetic algorithm according to claim 1, characterized in that, The hydrological parameter boundary group table includes parameter name, chromosome position number, upper limit value of parameter, lower limit value of parameter, and group number. The bi-group conditional migration table includes candidate vector number, error order, low contribution vector number, inflow item number, and outflow item number. The search state operator table includes elite vector number, stagnation algebra, parameter discrete value, feasible proportion value, and deterioration candidate vector number. The true error candidate sequence includes candidate vector number, runoff process error value, water volume deviation value, flood peak deviation value, and recession section deviation value. The hydrological model optimization parameter set includes runoff generation parameter value, evapotranspiration parameter value, infiltration parameter value, interflow parameter value, groundwater runoff parameter value, and confluence parameter value.

3. The method for optimizing hydrological model parameters based on genetic algorithm according to claim 1, characterized in that, The specific steps for obtaining the boundary group table of the hydrological parameters are as follows: Based on runoff generation parameters, evapotranspiration parameters, infiltration parameters, interflow parameters, groundwater runoff parameters, confluence parameters, upper limit of parameters, and lower limit of parameters, the six types of parameters are arranged in order, the parameter values ​​are compared with the boundary values, out-of-bounds markers are written, and a boundary verification vector set is generated. Based on the boundary verification vector set, out-of-bounds markers are filtered, reserved vector chromosome positions are bound, group numbers are assigned, and genetic master group numbers and exploration group numbers are written to obtain the hydrological parameter boundary group table.

4. The method for optimizing hydrological model parameters based on genetic algorithm according to claim 1, characterized in that, The specific steps to obtain the bigroup conditional migration table are as follows: Based on the hydrological parameter boundary group table, a genetic algorithm is used to read the candidate vector number of the genetic main group, the six types of parameter numbers, and the error value. The candidate vectors are arranged in ascending order of error, the same window comparison interval is defined, and the vector with lower error is selected window by window and the parent source number is recorded to generate the window winning vector set. Based on the windowed winning vector set, read the parent source number and the six types of parameter numbers, exchange the same parameter segments, limit the perturbation amplitude of the offspring, compare the upper and lower limits of parameters item by item, eliminate offspring vectors that exceed the limits and retain qualified vectors at the boundaries, and generate a candidate vector set for the offspring. Based on the candidate vector set of offspring, read the candidate vector error order, the number of consecutive low positions, and the elite vector distance value, mark the low contribution vectors, compare the candidate error of the exploration group with the quantile error of the main group, select the immigration item number and the immigration item number, and establish a bi-group conditional migration table.

5. The method for optimizing hydrological model parameters based on genetic algorithm according to claim 4, characterized in that, Using the genetic algorithm, the candidate vector numbers of the genetic main population, the six types of parameter positions, the error values, the upper limit of the parameters, and the lower limit of the parameters are read. The candidate vectors of the genetic main population are arranged in ascending order of error. Two to five candidate vectors are defined as the same window comparison interval. The runoff process error, water volume deviation, flood peak deviation, and recession section deviation are compared window by window. The candidate vector with the lowest error value is selected as the parent vector. The parent source number, parent group number, and parent chromosome position number are recorded. The parent isotopic parameter segments are exchanged according to the six types of parameter positions. The perturbation amplitude of the offspring is limited by the reciprocal of the parameter dimension. The offspring parameter values, upper limit of the parameters, and lower limit of the parameters are compared item by item. Off-limit offspring vectors are deleted, and qualified offspring vectors are retained. The parent source number, offspring vector number, error sequence, and boundary verification mark are summarized to generate the offspring candidate vector set.

6. The method for optimizing hydrological model parameters based on genetic algorithm according to claim 1, characterized in that, The specific steps to obtain the search state operator table are as follows: Based on the bi-group conditional migration table, the immigration item number, immigration item number, main group position number, and elite vector number are read. The immigration item is written into the immigration item position, the elite vector number and the corresponding six types of parameter values ​​are retained, the order of the genetic main group candidate vectors is rearranged, and the main group update vector set is generated. Based on the main group update vector set, read the continuous generation target error, the distribution values ​​of six types of parameters, the number of feasible vectors, the cumulative number of stagnant generations, calculate the parameter discrete values, calculate the feasible ratio values, write the search status markers according to the threshold interval, and generate a search status determination table. Based on the search state determination table, the simulated annealing algorithm is used to read the deterioration candidate vector number, parent error value, candidate error value, temperature value, and random number. The error difference is compared with the retention threshold, unqualified deterioration vectors are screened out, and the retention vector numbers are registered to obtain the search state operator table.

7. The method for optimizing hydrological model parameters based on genetic algorithm according to claim 6, characterized in that, Using the simulated annealing algorithm, the following are read: candidate deterioration vector number, parent error value, candidate error value, temperature value, random number, current generation number, and temperature segment number. The error difference is calculated by subtracting the parent error value from the candidate error value. When the candidate error value is less than the parent error value, the retained vector number is registered. When the candidate error value is greater than the parent error value, the exponential ratio between the error difference and the temperature value is calculated and used as the retention threshold. The random number is compared with the retention threshold. When the random number is less than the retention threshold, the retained vector number is registered. When the random number is greater than the retention threshold, the unqualified deteriorated vector is marked and filtered out. The retained vector number, temperature segment number, error difference, and retention threshold are summarized to obtain the search state operator table.

8. The method for optimizing hydrological model parameters based on genetic algorithm according to claim 1, characterized in that, The specific steps for generating the true error candidate sequence are as follows: Based on the search state operator table, read the temporary retained item number, check the parameter boundary and the constraints between parameters, mark the out-of-bounds vector and the constraint conflict vector, and generate a temporary vector verification table. Based on the temporary vector verification table, abnormal vectors are removed, and runoff process error, water volume deviation, flood peak deviation, and receding section deviation are calculated. Candidate positions are arranged to obtain the true error candidate sequence.

9. The method for optimizing hydrological model parameters based on genetic algorithm according to claim 1, characterized in that, The specific steps for obtaining the set of optimization parameters for the hydrological model are as follows: Based on the real error candidate sequence, the search state operator table is called, the upper limit algebra, the unimproved algebra, and the target error threshold are compared, a termination flag is written, and a termination judgment sequence table is generated. Based on the termination determination sequence list, the error sequence of runoff process, the sequence of water volume deviation, the sequence of flood peak deviation, and the sequence of water receding section deviation are read, and six types of parameter values ​​are mapped to obtain the hydrological model optimization parameter set.

10. A system for optimizing hydrological model parameters based on genetic algorithm, characterized in that, The method for optimizing hydrological model parameters based on genetic algorithm according to any one of claims 1-9, wherein the system comprises: Encoding and grouping module: Based on runoff generation parameters, evapotranspiration parameters, infiltration parameters, interflow parameters, groundwater runoff parameters, confluence parameters, upper limit of parameters, and lower limit of parameters, the module arranges the parameter positions, binds the chromosome positions, compares the boundaries, deletes out-of-bounds vectors, and obtains the hydrological parameter boundary group table; Interactive scheduling module: Based on the hydrological parameter boundary group table, a genetic algorithm is used to sort candidate errors, select window winning vectors, exchange parent sites, limit the perturbation amplitude of offspring, mark low contribution vectors, and pair immigration and emigration items to obtain a dual-group conditional migration table. Annealing screening module: Based on the bigroup conditional migration table, replace the migration items, retain the elite vectors, accumulate the stagnation algebras, compare the discrete values ​​of parameters, calculate the feasible proportion, and use the simulated annealing algorithm to screen the deteriorated candidate vectors to obtain the search state operator table; Runoff accounting module: Based on the search state operator table, it checks the boundaries and parameter constraints of temporary retained items, deletes abnormal vectors, calculates runoff process error, water volume deviation, flood peak deviation, and drainage section deviation, and obtains the true error candidate sequence; Termination output module: Based on the real error candidate sequence, call the search state operator table, compare the upper limit algebra, the unimproved algebra, and the target error threshold, map six types of parameters according to the error order, and obtain the hydrological model optimization parameter set.

Images